RE: Star's magnitude adjustments

["Gutzwiller, Michael" <mgutzwiller@lanvision.com>]

Arne wrote:

	
	Mike G. gave a description of his magnitude adjustment scheme.
Sounds
	fine on the surface.  A couple of questions:
	  If you have 8 regions, then there are 8 'constants' that are
used for
	the adjustment?  That is, each region uses a single value for
its
	adjustment and does not include a slope parameter?  Doesn't this
cause
	a discontinuity at the border between the regions?  Why not use
something
	like a quadratic/cubic/cubic spline instead and have it valid
over the
	entire image?

Looks like I forgot part of the description.  With 8 regions there are 9
adjustment values at the region boundaries.  The actual adjustment is
made by linear interpolation using the x position of the star and the
two nearest adjustment points.  I've looked at using a fit to the points
but it doesn't seem to buy you much.

	  The original flat vector is created from some nearby (in time)
photometric
	night?  How often do you update the flat vector?  How do you
handle a
	change in the flat vector (such as monitoring it to see when
changes occur)?

The flat vector is created by each site using whatever seems reasonable
to them.  For my own measurements I usually redo the flat vector
whenever I have a reasonable image to use.  For photometric nights
without a bright moon this is easy to do.  When there is a bright moon
or the night is non photometric it's harder and I may resort to using a
flat vector from a nearby night.  I haven't measured it but my guess is
that I have a flat vector for every two to three nights.

	  How many outliers are removed for a typical adjustment
calculation?  Is
	it a significant number compared to the ones left in the
calculation?

The median line fit doesn't actually throw outliers away.  It's just a
robust line fit in exactly the same way the median is a robust
measurement of the mean.

	  You mentioned earlier that you get different adjustments on a
nightly
	basis.  This sounds wrong to me.  Once the flatfield has been
calculated,
	a single adjustment should modify it sufficiently to make it
'flat'.  If
	you find a night-to-night variation in this adjustment, then I
don't
	understand what is happening.
	
I don't remember making the above statement.  I too would be upset if
the adjustment varied significantly unless the flat vector varied
significantly.  Most of the flat vectors I have examined seem quite
stable over long periods of time and I would expect the adjustment would
be stable also.  I am actively investigating the stability as part of
the beta testing.

Mike G.